Context-dependent reorganization of behavioral strategy during transfer from home-cage to head-fixed performance

Complex behavioral tasks increasingly rely on autonomous training paradigms that enable high-throughput learning in naturalistic settings. Such approaches offer a scalable route for preparing animals for subsequent head-fixed recordings required for high-resolution neural measurements. However, how animals adapt their behavioral strategies across these distinct contexts remains poorly understood. Here, we examine how mice trained autonomously in a group-housed home-cage system reorganize their behavior when transitioned to a head-fixed version of the same delayed-response task. Although overall success rates were comparable across contexts, mice exhibited a striking shift in behavioral policy. Freely behaving home-cage performance was characterized by high engagement and premature responses, whereas head-fixed performance showed a progressive emergence of selective, cue-intensity-dependent responding accompanied by increased omission errors. This selective strategy was strongly associated with improved performance and sustained engagement under head-fixed conditions, reflecting a trade-off between sensitivity to low-intensity cues and suppression of premature actions. Importantly, selectivity did not arise immediately upon head fixation but developed gradually with experience across sessions, indicating learned adaptation rather than a spontaneous response to restraint. Our findings demonstrate that behavioral policy is not solely determined by task rules or sensory demands, but is strongly shaped by task architecture, temporal constraints, and opportunity cost. These results highlight the need to account for context-dependent strategy shifts when interpreting neural activity following transfer from autonomous training to controlled recording environments, and establish a framework for studying adaptive behavior across experimental contexts.